Demonstration · Chapter Four

One slider
removes
eighty-six
percent.

The trouble is that nobody uses one slider. Set the ones you would actually set, and watch what each honest sentence costs.

Settings with a number

Region
i
Where this comes from
The 100 largest US metros, with 2024 population estimates from the Census Bureau. People in range are counted from real geography: your metro fills in as the radius grows, and every other metro whose centre falls inside the circle is added. So the answer depends on where you are, not only how far you will drive. Sixty miles is the default because it is roughly what people treat as drivable for a first date.

Phoenix at sixty miles comes to about five million, which is the figure Chapter Four uses.

Assumption
5.4% of any metro is men aged 28–35, held constant across all 100. That is calibrated to Phoenix and will be wrong for a college town or a retirement coast. Population is also assumed to concentrate in a metro core rather than spread evenly, and a neighbouring metro is counted by how close its centre sits to the edge of your circle rather than carved up precisely.
 270,000
60 miles

5.0 million people in range · 270,000 men aged 28–35

Looking for
i
Where this comes from
This sets which population the filters run against. The worked example in Chapter Four follows a woman filtering men, because that is the direction the height data describes.
 Baseline

Age
i
Where this comes from
US Census 2024 population estimates, by single year of age and sex. The band you set is priced against the whole adult population of the metro, so widening it genuinely adds people rather than just relabelling them.

Chapter Four uses 28–35, which is 11.4% of American men.

Assumption
Age is treated as a clean gate, which on a dating app it very nearly is — it is the one filter nobody can present well, argue with, or improve. Everyone outside the band simply never appears. The real distribution of who is available also shifts with age, and that is ignored here.
 11%

28 to 35

28
35
Height
i
Where this comes from
14.5% of American men reach six feet (CDC anthropometric data). A 2022 Wall Street Journal report quoted a former Bumble product manager saying most women on the app set this filter at 6′0″ or above. Bumble has never released its own figures.
 −86%

6′0″ and above

Income
i
Where this comes from
Individual earnings for US men aged 28–35, from Census income tables. Applied after the height filter, which is why the combined figure is higher than multiplying the two.

Assumption
Earnings are modelled as a lognormal curve with a median near $55,000, tuned so that six figures lands on the roughly 15% the Census tables show. Real income has a fatter top tail than any smooth curve, so the figures above $150,000 are approximate.
 −85%

$100,000 and above

Education
i
Where this comes from
Educational attainment for US men aged 25–34. Census and ACS put bachelor’s or higher at about 36% for this group, against 46% for women the same age.

Like the sliders above it, this figure is what the setting removes on its own. Degrees and income travel together, so stacking the two costs you far less than multiplying them suggests — which is what the box below the grid works through.

Assumption
The master’s figure is estimated at 9%. Census publishes master’s-or-higher for men 25 and over at roughly 10%, and the 25–34 group sits below that because many have yet to finish. The associate and some-college stops are interpolated the same way.
 −64%

Bachelor’s degree or higher

Currently single
i
Where this comes from
Roughly a third of men in this age band are unpartnered, from Census marital-status data. This is the filter people forget to count, and it removes more than height does.
 −66%

Unmarried and not partnered

Attractiveness
i
Assumption
This is a definition, not a measurement. Splitting a population into below / average / above as 25 / 50 / 25 is an assumption I am making so the control does something — attractiveness has no agreed scale, and ratings differ sharply between raters. Treat the number as illustrative.
 −75%

Above average

Body type
i
Assumption
These four shares are mine, not anyone’s data. Self-reported body type is systematically optimistic, the categories are not standardised across apps, and they are applied unevenly by gender — curvy is rarely offered on a men’s profile at all. Treat the split as illustrative.
 −20%

Athletic

Religion
i
Where this comes from
Shares of US adults by tradition, from the Pew Religious Landscape Study. This is where the order-of-magnitude problem shows itself: filtering to Christian keeps six in ten, filtering to Hindu keeps one in a hundred. Same control, same single sentence, wildly different cost. And these count people who identify with a tradition — requiring someone who actually practises cuts every figure again, by an amount nobody has measured well.
 −38%

Christian

Politics
i
Where this comes from
Self-reported ideology among US adults, from Gallup’s long-running trend. Roughly a quarter liberal, with moderates and conservatives splitting the rest. Partisan composition swings by tens of points between metros, so a national figure understates how much this filter costs in some places and overstates it in others.
 −64%

Liberal

Settings without one

Wants children
i
Assumption
Survey answers on intent vary widely by age, region and question wording, and stated intent is a poor predictor of what people actually do. No reliable single figure exists, so this one shows direction only.
 No number

On roughly your timeline

Some preferences have no honest number at all. Stated intent is one of them — it shifts with how the question is asked and predicts behaviour poorly. There is only the direction, and the direction is always down.

EthnicitySwitch removed

For years you could filter by ethnicity — no admission required, nothing to defend, nobody’s face to watch while you did it. Grindr removed the option in 2020. The preference did not go anywhere. The company stopped handing people the switch.

6,000 men left

From 270,000. Two settings so far.

Each square is 1,000 men · yellow is who is left

The maths, done honestly

Multiply the percentages, the way the internet does
700
Corrected: height, income and education travel together
1,500

Most versions of this calculation multiply independent odds and arrive at a number roughly half the real one. Correcting it in the generous direction still leaves fifteen hundred men in a metro area of five million.

Four settings in

Everyone who clears four
ordinary preferences would fit
inside a high school gymnasium

Nobody chose that. Every one of those settings is defensible by itself, and the person setting the last one has no way to feel what the others already did.

Four settings got you to a gymnasium. Eight gets you to a room.

The hand

Somebody built the switch

A 2025 YouGov survey found that 55 percent of women and 56 percent of men support letting dating apps filter by height at all. Near-identical approval for a feature that, once switched on, works almost entirely in one direction.

Here is the part its users rarely anticipate. When most women set the same threshold, the pool of eligible men gets smaller for each of them — and all of that attention lands on the same few men. A small number above the line receive nearly all of it, which changes how they behave. The much larger group below it becomes invisible, whatever else they bring. The filter was supposed to make the search efficient. At scale it built a market that frustrates nearly everyone.

Good news

This is not how
people actually pair

In observed relationships the distribution of couples by height looks nothing like the distribution of height filters. Most men in real couples are under six feet.

The preference, made explicit and consequence-free, overshoots what people actually do. In person they make compromises, fall for someone’s laugh, get disarmed by someone’s attention. Ask directly rather than reading it off a settings screen and the asymmetry is still there, at a fraction of the size.

So the useful move is to set fewer sliders and meet more people. The settings screen is a worse predictor of who you will love than almost anything else you could use.

Read Chapter Four